analysis of the influence of pci planning on the …webpersonal.uma.es/~toril/files/2017 wpc pci...
TRANSCRIPT
Analysis of the Influence of PCI Planning on the PhysicalUplink Control Channel in LTE
R. Acedo-Hernandez1 • M. Toril1 • S. Luna-Ramırez1 •
J. A. Fernandez-Segovia1 • C. Ubeda2
� Springer Science+Business Media, LLC 2017
Abstract In Long Term Evolution (LTE) systems, the physical cell identity (PCI)
assigned to a cell during network planning determines the set of sequences used by sub-
scribers as de-modulation reference signals (DM RS) in the uplink (UL). A proper
assignment of PCIs must prevent neighbor cells from using the same DM RS to avoid
interference problems, thus ensuring adequate performance both in control and user data
channels in the uplink. In this paper, a comprehensive analysis is carried out to quantify the
impact of PCI planning on the performance of physical uplink control channel (PUCCH) in
LTE. First, a novel analytical model that reflects the influence of PCI planning on inter-
ference and outage probability due to DM RS collisions in PUCCH is presented. Based on
this model, PUCCH performance with several classical PCI planning schemes is evaluated
with a static system-level simulator implementing a real network scenario. Simulated cases
cover different PUCCH frame formats, different UL power control (PC) schemes and both
regular and irregular traffic scenarios. Results show that a PCI plan constructed solely
based on avoiding PCI collision/confusion and reference signals collisions in the downlink
achieves near-optimal performance in terms of DM RS collisions in PUCCH, but, how-
ever, in extreme cases, DM RS collisions caused by neighbor cells sharing the same
sequence might degrade PUCCH performance significantly.
Keywords Mobile network � Long Term Evolution � Planning �Physical cell identity � Demodulation reference signal � Uplink
& R. [email protected]
1 Departamento Ingenierıa de Comunicaciones, Universidad de Malaga, Malaga, Spain
2 Ericsson, 28045 Madrid, Spain
123
Wireless Pers CommunDOI 10.1007/s11277-017-4887-7
1 Introduction
Mobile networks are continuously growing in size and complexity, which has made mobile
network management a very time-consuming task. To reduce the workload of system
management, mobile operators try to automate all the tasks in the network life cycle. For
this reason, a great effort has been made by the industry and standardization bodies to
define and develop automatic network management procedures [24]. As a result, the
recently deployed Long Term Evolution (LTE) standard is well known for its self-orga-
nizing networks (SON) capabilities [17, 26]. In this context, SON refers to the capability of
the network to perform self-planning, self-optimization and self-healing with minimal
human intervention.
The assignment of physical cell identity (PCI) in LTE has been identified as one of the
most relevant SON use cases [3, 13]. The PCI is a code used to identify a base station in
user mobility procedures [15]. Such a code is assigned during network planning in the PCI
plan. The number of PCI values is limited to 504, which entails repeating PCI values across
the network if there exists more than 504 cells. An improper PCI assignment in a cell may
negatively affect uplink (UL) and downlink (DL) performance in that cell and its neigh-
bors. For instance, if two neighbor cells share the same PCI, a user equipment (UE) cannot
correctly identify its serving cell (referred to as collision problem). Likewise, when a
serving cell has two neighbors with the same PCI, the UE cannot identify the target cell
during a handover (referred to as confusion problem), and a handover failure occurs.
In the literature, most research efforts have focused on designing automatic PCI plan-
ning algorithms that avoid collision and confusion problems caused by PCI reuse. For this
purpose, PCI planning is formulated as a graph coloring problem, which is solved by
graph-theoretic algorithms [2, 4, 20, 32]. Initial proposals considered centralized algo-
rithms, conceived for the network design stage [10, 23]. More recent proposals suggest
distributed approaches for plug-and-play operation of eNodeBs (eNB) [11, 22, 30].
Later studies have extended the analysis of PCI planning to heterogeneous LTE net-
works [21, 25, 31, 33, 34].
But, at the same time, the PCI also defines the location of cell-specific reference signals
(CRS) in the frequency domains in DL [27]. An inadequate PCI plan can lead to CRS
collisions between neighbor cells, causing that connection quality is not estimated correctly
and the efficiency of DL data transmission is reduced [35]. Some studies [1, 29] have
evaluated the influence of PCI planning on DL CRS collisions. In [29], a heuristic PCI
planning algorithm is proposed to keep PCI reuse distance within certain limits while still
avoiding DL CRS collisions in sectors of the same site. A more detailed analysis of the
impact of PCI planning on DL throughput performance due to CRS collisions is presented
in [1].
However, to the authors knowledge, no study has quantified the impact of PCI planning
on UL performance. The PCI plan has an impact on UL because it defines the set of
demodulation reference signals (DM RS) used in a cell for channel estimation and coherent
demodulation in the physical uplink shared channel (PUSCH) and physical uplink control
channel (PUCCH) [27]. If two users of different cells use the same DM RS, the decoding
process in PUCCH or PUSCH is degraded, which might cause that the base station could
not identify the user correctly. As it has been said, currently, operators design their PCI
plans to avoid problems on the DL. This is due to the fact that, in PUSCH, the assignment
of DM RSs can be decoupled from PCI values by using a shift parameter so that DM RS
collisions between neighbor cells can be avoided easily [27]. However, this is not the case
for PUCCH, where PCI planning might have an influence on control channel performance.
R. Acedo-Hernandez et al.
123
This issue will be of the utmost importance in future 5G networks, where uplink trans-
mission is key for Internet-of-thing applications [16].
In this work, a comprehensive performance analysis is carried out to check the influence
of PCI planning on PUCCH performance due to DM RS collisions. The core of the analysis
is a novel analytical model of the influence of PCI planning on PUCCH failure probability.
The proposed PUCCH performance model is included in a static system-level simulator
implementing a real scenario, with which the impact of PCI planning on PUCCH failure
probability is quantified. The main contributions of this work are: (a) a PUCCH perfor-
mance model that can be used to evaluate the impact of PCI planning on LTE UL signaling
performance, and (b) a performance analysis of PUCCH failure probability due to DM RS
collisions for different PCI planning approaches and network configurations in a realistic
scenario. The rest of the paper is organized as follows. Section 2 formulates the influence
of PCI planning on DM RS collisions. Then, Sect. 3 explains the PUCCH performance
model developed in this work. Section 4 presents the experiments carried out to assess the
different PCI planning approaches. Finally, Sect. 5 presents the conclusions of this work.
2 Problem Formulation
The aim of this section is to clarify the conditions when two PUCCH users collide due to
their DM RS, and how these conditions depend on the PCI plan. First, a description of
PUCCH structure and frame formats is given. Second, system-level techniques associated
to PUCCH reference signals are briefly discussed. These techniques include group
sequence, cyclic shift, code multiplexing and hopping. The focus here is to derive how
many users can be handled simultaneously in the PUCCH of a cell. Finally, the conditions
when two PUCCH users collide due to their DM RS are presented.
2.1 PUCCH Structure
Figure 1 shows time and frequency resources allocated to PUCCH. Each UE is assigned a
PUCCH region, consisting of one resource block (RB) and one sub-frame (1 ms). Each
PUCCH region alternates its first slot in one edge of the system bandwidth and the second
one at the opposite edge (up and down in the figure) to ensure frequency diversity.
The exact number of RBs reserved for PUCCH is determined on a cell basis during
network planning depending on system bandwidth [27].
In PUCCH, different UEs can be separated by frequency division multiplexing (FDM)
(i.e., using different RBs). Additionally, several UEs can be separated in the same PUCCH
region by code division multiplexing (CDM) (i.e., using different codes), as detailed in
Sect. 2.3.2.
2.2 PUCCH Formats
Table 1 presents the different PUCCH frame formats. Briefly, format 1 is used for
scheduling requests (SR), formats 1a/1b for hybrid automatic repeat request (HARQ)
ACK/NACK transmissions and formats 2/2a/2b for channel quality indicator (CQI)
transmissions. The table also presents the multiplexing capability of each format, reflecting
the maximum number of UEs per RB, which is explained later.
Analysis of the Influence of PCI Planning on the…
123
Figure 2 shows how PUCCH transmissions are grouped by frame format in frequency
subregions comprising one or more RBs. PUCCH formats 2/2a/2b (devoted to CQI
transmission) are mapped on the edges of system bandwidth, while PUCCH SR/HARQ
format 1/1a/1b (for SR/ACK/NACK transmission) are transmitted on the innermost RBs.
The index m in the figure denotes a specific PUCCH region. Identifiers #0, #1 and #2 are
used to distinguish between frames of equal format.
Figure 3 represents the time structure of the different PUCCH frame formats in a
symbol level. DM RS denotes Reference Signal, LB Long Block and CP Code Prefix.
Fig. 3a shows that, in formats 1/1a/1b, shaded symbols #2, #3 and #4 are used for DM RS.
In contrast, Fig. 3b shows that symbols #1 and #5 are used for DM RS in formats 2/2a/2b.
1 sub-frame (1ms)
12 subcarriers (RB)
Slot 0 (0.5 ms) Slot 1 (0.5 ms)
PUSCHN RBs SystemBandwidth
PUCCH
PUCCH
PUCCH region #0
PUCCH region #0
Time
Frequency
PUCCH region #1
PUCCH region #1
PUCCH region #2
PUCCH region #4
PUCCH region #3
PUCCH region #3PUCCH region #2
PUCCH region #4
PUCCH region #5
PUCCH region #5
Fig. 1 PUCCH structure
Table 1 PUCCH frame formats [18]
Format Uplink control information Multiplexingcapacity[#UE/RB]
1 Scheduling request (SR) 36, 18, 12
1a 1-bit HARQ ACK/NACK with/without SR 36, 18, 12
1b 2-bit HARQ ACK/NACK with/without SR 36, 18, 12
2 CQI (20 coded bit) 12, 6, 4
2a CQI and 1-bit HARQ ACK/NACK (20 ? 1 coded bits) 12, 6, 4
2b CQI and 2-bit HARQ ACK/NACK (20 ? 2 coded bits) 12, 6, 4
R. Acedo-Hernandez et al.
123
2.3 PUCCH Reference Signals
In [15], two types of UL RS are defined:
• Demodulation RS (DM RS), used for channel estimation in coherent demodulation in
PUSCH and PUCCH, and
• Sounding RS (SRS), not associated with any specific data or control channel, and used
to estimate channel quality to enable frequency-selective scheduling in the UL.
All UL RSs are based on Zadoff-Chu (ZC) or QPSK-alphabet sequences, designed to
ensure low cross-correlation between sequences [5]. RSs must occupy the full bandwidth
assigned to the user for data transmission. This forces that the RS sequence length in
1/1a/1b #5 1/1a/1b #4
1/1a/1b #4 1/1a/1b #5
1/1a/1b + /2a/2b #2 1/1a/1b #3
2/2a/2b #0 2/2a/2b #1
1 sub-frame (1ms)
12 subcarriers (RB)
Slot 0 (0.5 ms) Slot 1 (0.5 ms)
PUSCHN RBs systembandwidth
PUCCH
PUCCH
2/2a/2b #0
Time
Frequency
2/2a/2b #11/1a/1b #3 1/1a/1b + 2/2a/2b #2
Fig. 2 Frequency allocation of different PUCCH frame formats
CP LB #0 C
P LB #1 CP RS LB C
P RS LB CP RS LB C
P LB #5 CP LB #6
1 slot (0.5 ms)1 symbol
CP LB #0 C
P RS LB CP LB #2 C
P LB #3 CP LB #4 C
P RS LB CP LB #6
1 slot (0.5 ms)
1 symbol
(a)
(b)
Fig. 3 Time structure of different PUCCH frame formats (normal cyclic prefix). a Format 1/1a/1b.b Format 2/2a/2b
Analysis of the Influence of PCI Planning on the…
123
symbols, Nm, is equal to the number of assigned sub-carriers, MRSsc , which is 12 times the
number of assigned RBs to the UL transmission, m, as
Nm ¼ MRSsc ¼ m � 12 1�m�NUL
RB ; ð1Þ
where NULRB is the total UL system bandwidth in RBs.
2.3.1 Sequence Group
As DM RSs must have the same bandwidth as data, at least one RS sequence of each length
Nm is needed per cell to support all possible RB allocation sizes to users in the UL of a cell.
The number of possible base RS sequences of each length Nm varies with the number of
assigned RBs, m. The minimum number of sequences for a certain length is 30, obtained
for m 2 f1; 2; 3g [27]. Thus, the complete set of available sequences is divided into 30 non-
overlapping sequence groups, each including at least one base sequence of all lengths
(specifically, 1 for m 2 ½1; 5� and 2 of larger lengths). During network planning, every cell
is assigned one of these sequence groups, denoted by the subgroup index u, where
u 2 f0; 1; . . .; 29g.
2.3.2 Cyclic Shift and User Multiplexing in PUCCH
In PUCCH, several simultaneous users may use the same RB. All these users are assigned
the same number of RBs (i.e., 1), and thus have the same base sequence length (i.e., 12). To
support multiple users in a RB, a cyclic time shift, a, can be applied to the base sequen-
ce(s). Ideally, cyclic shifts of the same base sequence are fully orthogonal, unlike any two
base sequences (i.e., users in different cells), which show non-zero cross-correlation. This
property makes cyclic shifts the best option to separate users transmitting on the same set
of sub-carriers, known as multi-user multiple-input multiple-output (MU-MIMO).
For a DM RS, there are 12 possible cyclic shifts, but some of them may not be available
in a cell. The circular distance parameter, Dshift, controls the number of available cyclic
shifts in a cell. Possible values of Dshift are 1, 2 and 3, corresponding to 12, 6 or 4 available
cyclic shifts, respectively.
In formats 1/1a/1b frames (used for SR/ACK/NACK feedback), time-domain spreading
codes can also be used for UE multiplexing in addition to the 12 available cyclic time
shifts. The number of available spreading codes is limited by the number of RS (3, in this
format, as shown in Fig. 3). Thus, up to 36 (¼ 12 � 3) UEs can be orthogonally multi-
plexed on the same PUCCH RB in formats 1/1a/1b, as shown in Table 1. In formats 2/2a/
2b frames (used for CQI reporting), time-domain spreading codes are not available, so that
only up to 12 UEs can be multiplexed on the same RB.
2.3.3 Randomization by Hopping
The assignment of a sequence group to a cell and cyclic shift to a user in LTE can be fixed
or, alternatively, may vary dynamically with time following a hopping pattern. Sequence-
group and cyclic-shift hopping randomize inter-cell interference, which would otherwise
be concentrated on a single user all the time. The price to be paid is the inability to
optimize inter-cell interference by a careful RS planning across the network.
If sequence-group hopping is enabled, the sequence group u assigned to a cell in each
time slot is determined by a sequence-group hopping pattern, fgh, and a sequence-group
shift offset, fss, as
R. Acedo-Hernandez et al.
123
uðnsÞ ¼ ðfghðnsÞ þ fssÞmod 30; ð2Þ
where ns is the current time slot, fgh is a pseudo-random sequence taken from a set of 17,
and fss is a number between 0 and 29. Thus, there are 17 unique sequence-group hopping
patterns, each of which can be shifted by 30 different sequence-group shifts offsets. The
sequence-group hopping pattern index in a cell c, cinit, is given by the PCI of the cell,
PCI(c), as
cinitðcÞ ¼jPCIðcÞ
30
k2 ½0; 16� ð3Þ
where PCI(c) 2 f0; . . .; 503g [15]. Thus, up to 30 consecutive PCI values can have the
same fgh.
The sequence-group hopping pattern index, fgh, in a cell is the same for PUCCH and
PUSCH, but the sequence-group shift offset, fss, may be different for PUCCH and PUSCH.
For PUCCH,
fss ¼ PCIðcÞmod 30; ð4Þ
whereas, for PUSCH,
fss ¼ PCIðcÞmod 30þ Dss; ð5Þ
where Dss is a parameter for decoupling the assignment of sequence group from the
assignment of PCI, Dss 2 f0; . . .; 29g. If sequence-group hopping is disabled, uðnsÞ ¼ fss.
Cyclic shift hopping is added on top of sequence-group hopping. In PUSCH, cyclic shift
hops every slot, whereas, in PUCCH, cyclic shift hops every SC-FDMA symbol.
2.4 PUCCH Collision
Two PUCCH users collide when they use the same base RS sequence and cyclic shift in a
symbol period. Ideally, orthogonal cyclic shifts avoid collisions between users in the same
cell so that collisions only take place between users of different cells. So, collisions
between users of different cells happen when: (a) the cells serving the two users must have
the same base RS sequence group index, u (which depends on their PCI values), (b) the two
users must be assigned to the same PUCCH RB (i.e., the same PUCCH region), and (c) the
two users must have the same cyclic shift and code (when available). Should RS sequences
be fully orthogonal between them, the three previous conditions would be necessary
conditions for collision.
However, RS sequences are not perfectly orthogonal, which is especially true for the
sequences of limited length used in PUCCH [8]. Thus, partial collision is also possible
between PUCCH users with different RS sequence group and cyclic shift. The impact of
these partial collisions depends on the orthogonality factor between the specific RS
sequences assigned to users on a symbol basis.
3 PUCCH Performance Model
Providing a simple and accurate PUCCH performance model is essential to assess the
quality of a PCI plan without the need for time-consuming simulations. In this work, a
methodology for estimating PUCCH failure probability due to DM RS collisions with a
Analysis of the Influence of PCI Planning on the…
123
static system-level simulator is described. In the proposed performance model, the PUCCH
failure probability obtained by a given PCI plan is estimated based on the DM RS collision
probability and PUCCH interference level. The model covers different PUCCH formats
(i.e., format 1/1a/1b and 2/2a/2b) and different PCI configurations (with and without
hopping).
3.1 Calculation of PUCCH Interference Level
A first step in the model is to derive the received interference level in PUCCH. It is
assumed that hopping techniques ensure that the interference averaged over 2 slots (i.e., 14
SC-FDMA symbols with normal cyclic prefix) is nearly constant. Thus, interference can be
characterized by its average value.
The average PUCCH interference level (in natural units) is computed on a cell basis as
the sum of intracell and intercell interference,
IðcÞ ¼ IintraðcÞ þXc0 6¼c
Iinterðc; c0Þ; ð6Þ
where IintraðcÞ is the average received interference level due to users in the same cell c, and
Iinterðc; c0Þ is the average received interference level in cell c from users in surrounding
cells c0.Both interference terms are defined as the product of the number of interfering users, a
potential received signal level strength and an interference protection factor, as
IintraðcÞ ¼ NusintraðcÞ � Prxðc; cÞ � Xintra; ð7Þ
Iinterðc; c0Þ ¼Xc0 6¼c
Nusinter ðc0Þ � Prxðc; c0Þ � Xinterðc; c0Þ ; ð8Þ
where NusintraðcÞ and Nusinter ðc0Þ is the average number of interfering users from cell c and
surrounding cells c0, Prxðc; cÞ and Prxðc; c0Þ is the average received signal level in cell c
from users in the same cell and in surrounding cells, respectively, and Xintra and Xinterðc; c0Þare interference protection factors due to precoding (i.e., cyclic shift and code division
multiplexing). It is assumed that the code plan is independent between users. As shown
later, the number of users depends on PUCCH traffic intensity. Likewise, the average
received signal level depends on the UL power control scheme and the protection factor
depends on the PCI plan and the code utilization ratio.
3.1.1 Number of Interfering Users
The number of interfering users in (7) and (8) depends on the traffic intensity per PUCCH
region. The average number of interfering users served by the same cell can be computed
from the probability of having additional users (given that there is at least one user), as
NusintraðcÞ ¼ E½nusðcÞ� 2 j nusðcÞ� 1� ¼XNusmax
i¼2
i � pðnusðc0Þ ¼ iÞ
pðnusðcÞ� 1Þ ;ð9Þ
where nusðcÞ is a random variable showing the number of users in cell c, ranging from 0 to
Nusmax , Nusmax is the maximum number of users in the cell, given by the number of cyclic
shifts and codes, and pðnsðcÞÞ is the probability of having nusðcÞ active users in cell c in the
R. Acedo-Hernandez et al.
123
considered PUCCH region. In (9), note that the probability of having intracell interferers is
conditioned to the fact that there is already a user in the cell. Likewise, the average number
of interfering users from other cells c0 is computed as
Nusinter ðc0Þ ¼ E½nusðc0Þ� ¼XNusmax
nusðc0Þ¼1
nusðcÞ � pðnusðc0ÞÞ ð10Þ
For tractability, it is assumed here that each PUCCH region can be treated independently
(i.e., traffic intensity is described on per-region basis and interference only comes from
users in the same PUCCH region). In each PUCCH region, the request arrival process per
cell is modeled as a Poisson process with arrival rate kðcÞ, and the service time of each
request is one subframe (i.e. 1 ms). With these assumptions, the probability of having nrrequests per subframe depends on the arrival rate in the considered PUCCH region as:
pðnrÞ ¼e�ðkTÞðkTÞnr
nr!; ð11Þ
where k is the PUCCH request arrival rate per RB (i.e., per PUCCH region) and T is the
subframe duration (i.e., 1 ms). Note that nr ¼ k � T is the average number of PUCCH
requests per subframe/region in the cell at any time.
The maximum number of requests that can be carried per PUCCH subregion in a
subframe is limited by PUCCH capacity. By taking this limit into account, the probability
of serving nus users in a subframe of a PUCCH region can be expressed as:
pðnusðcÞ ¼ 0Þ ¼ e�nr
pðnusðcÞ ¼ 1Þ ¼ e�nr � nr
pðnusðcÞ ¼ 2Þ ¼ e�nr � nr22!
..
.
pðnusðcÞ ¼ NusmaxÞ ¼ pðnr ¼ NusmaxÞ þ pðnr ¼ Nusmax þ 1Þ þ � � � þ pðnr ! 1Þ:
ð12Þ
The value of Nusmax depends on the PUCCH format used in the considered PUCCH region
and the value of the Dshift parameter. A typical configuration is Dshift ¼ 2, resulting in
Nusmax ¼ 6 for format 2/2a/2b and Nusmax ¼ 18 for format 1/1a/1b.
3.1.2 Average Potential Interference Level
In (7) and (8), the potential average interference from cell c0 to cell c, Prxðc; c0Þ, depends onthe UL transmit power and pathloss, as
Prxðc; c0Þ ¼ EPtxðuÞPLðc; uÞ
��� SðuÞ ¼ c0� �
; ð13Þ
where PtxðuÞ is the transmit power of user u in cell c0, PL(c, u) is the pathloss (including
antenna gains and slow fading) of user u to cell c, and S(u) is the cell serving user u (i.e.,
c0). PtxðuÞ is defined by the UL power control (PC) scheme, as explained later.
Analysis of the Influence of PCI Planning on the…
123
3.1.3 Interference Protection Factors
The interference protection factors in (7) and (8), Xintra and Xinter , mainly depend on
sequence correlation properties. These values have been defined according to the
orthogonality conditions between users. These conditions are detailed below. Specifically,
Xintra ¼ acs�cdmintra; ð14Þ
where acs�cdmintrais the orthogonality factor between time-aligned versions of the same base
sequence with different precoding (i.e., cyclic shift and multiplexing code). In contrast, the
intercell interference protection factor, Xinter , also depends on the PCI plan and the code
utilization ratio.
When sequence-group hopping is disabled, DM RS collisions occur between cells
planned with the same PCImod 30. Thus,
Xinterðc; c0Þ ¼acs�cdminter
ðc; c0Þ if PCIðcÞmod 30 ¼ PCIðc0Þmod 30 ;
abs otherwise;
�ð15Þ
where abs is the orthogonality factor between different base sequences, and acs�cdminteris the
orthogonality factor between versions of the same base sequence with different precoding
(i.e., cyclic shift and multiplexing code). Thus, acs�cdminteris defined as,
acs�cdminterðc; c0Þ ¼
1
Nusmax
� acs�cdmintraif c and c0arecosited;
1 otherwise:
8<: ð16Þ
Note that, for co-sited cells, the product E½nusðc0Þ� � 1Nusmax
, obtained in 8 is the average code
utilization ratio in the interfering cell c0. The lower this ratio, the lower the average
intercell interference generated by cell c0 in co-sited cell. Such a protection does not exist
between non-cosited cells.
With sequence-group hopping enabled, base sequence collisions only occur between
cells with different hopping lists every 1 out of 30 slots. Thus,
Xinterðc; c0Þ ¼abs if
jPCIðcÞ30
k¼
jPCIðc0Þ30
k;
29abs þ 1
30otherwise;
8>><>>:
ð17Þ
In the previous expression, it is assumed that co-sited cells always share the same base
sequence hopping list, so that precoding is not effective between different cells.
Ideally, different base sequences are fully orthogonal, and the same holds for precoded
versions of the same base sequence (i.e., abs ¼ acs�cdminter¼ 0). In practice, some corre-
lation exists between sequences (i.e., abs; acs�cdminter2 ð0; 1�) [6, 8].
3.2 Calculation of PUCCH Outage Probability
The PUCCH outage probability is the probability that the PUCCH SINR averaged over 2
slots is below a certain threshold. The average SINR (in dB) for a user u is estimated by
subtracting the average desired signal and received interference levels (in logarithmic
units) as
R. Acedo-Hernandez et al.
123
SINRðuÞ ¼ PrxðuÞ � ðIðSðuÞÞ þ N0Þ
¼ PrxðuÞ � ðIintraðSðuÞÞ þX
c0 6¼SðuÞIinterðSðuÞ; c0Þ þ N0Þ; ð18Þ
where SINR(u) is the average PUCCH SINR of user u, PrxðuÞ is the average desired signal
level received at the base station, fixed by the UL PC scheme, and IðSðuÞÞ is the average
PUCCH interference level in the cell serving user u, S(u), which is shared by all users in
the same PUCCH region of the cell.
To compute interference levels, the transmit power of each PUCCH user, PtxðuÞ, isobtained from the Uplink Power Control (ULPC) formula [12]:
PtxðuÞ ¼ min PtxðuÞmax;P0;PUCCH þ a � PLðc0; uÞ þ dmcs þ f ðDiÞ� �
; ð19Þ
where Ptxmax is the maximum transmit power of the terminal, P0PUCCH is the nominal
received power parameter in PUCCH, PLðc0; uÞ is the pathloss from user u to cell c0, a is
the path loss compensation factor, dmcs is a MCS dependent offset and f ðDiÞ is a closed
loop correction factor. In this work, two different UL PC schemes are considered, namely
open-loop PC and fast closed-loop PC. For open-loop PC, f ðDiÞ ¼ 0, while for closed-loop
PC, the UE is able to adjust the uplink transmit power in accordance with the closed-loop
correction values. In this case, the uplink receiver at the eNB estimates the SINR of the
received signal and compares it with the desired SINR target value. If the received SINR is
below the SINR target, a command in sent to the UE to request for an increase in the
transmitter power. Otherwise, a decrease will be required in transmitter power.
Once the average PUCCH SINR value is obtained for each user, the PUCCH outage
probability in a cell is estimated by the share of users that do not satisfy the minimum
SINR threshold, SINRth, as
PoutageðcÞ ¼ pðSINRðuÞ� SINRth j SðuÞ ¼ cÞ: ð20Þ
4 Performance Analysis
The above-described model is included in a system-level simulator to assess PUCCH
performance with different PCI planning approaches in a realistic scenario. For clarity, the
simulation set-up is first introduced and results are then presented.
4.1 Simulation Set-up
Figure 4 shows the simulated real scenario, consisting of 699 base stations covering a large
geographical area of 2900 km2. The area includes a mixture of environments, with densely
populated areas and rural areas, whose intersites distances ranges from 1.5 to 14 km.
The previous scenario is implemented in a static system-level simulator [9]. Table 2
shows the main parameters of the simulation tool.
The scenario is divided into a grid of points, representing potential user locations, for
which propagation losses to each base station are calculated with COST-231 Hata prop-
agation model [7]. Log-normal slow fading is also considered. From this data, cell service
areas are calculated. PUCCH traffic is modeled with a Poisson arrival process with total
arrival rate k, which is then evenly or unevenly distributed among the different cells,
Analysis of the Influence of PCI Planning on the…
123
depending on the selected spatial traffic distribution. In each cell, PUCCH attempts are
evenly distributed across user locations in the service area. Following vendor suggestion,
Dshift ¼ 2. Thus, the maximum number of users in each PUCCH region is 18 or 6 for
format frame type 1/1a/1b and 2/2a/2b, respectively. Transmit power for each user location
is defined by the selected PC scheme. Both open-loop and fast closed-loop PC are con-
sidered. Following common practice, the nominal power parameter, P0;PUCCH , is fixed to -
120 dBm. Potential received signal levels are computed with propagation data. Then,
PUCCH interference level, experienced SINR and outage probability are calculated on per-
user basis with the model described in Sect. 3.
To check the impact of model assumptions, 4 use cases are defined:
a. Case 1 (Ideal case) Fast closed-loop PC, abs ¼ acs�cdmintra¼ 0, acs�cdminter
¼ 0 for
cosited cells and 1 otherwise, uniform spatial user distribution.
b. Case 2 (Non-ideal PC) Similar to case 1, but with open-loop PC.
c. Case 3 (Real partial orthogonality) Fast closed-loop PC, abs ¼ 0:3 [8], acs�cdmintra¼
0:2 [6], acs�cdminter¼ 0:2
Nusmaxfor cosited cells and 1 for non-cosited cells, and uniform
spatial user distribution.
d. Case 4 (Real irregular distribution) Similar to case 3, but with irregular spatial user
distribution. In this case, the share of users in each cell is derived from PUCCH
attempt statistics taken from the live network.
In all use cases, 4 different PCI planning methods are tested:
a. UL-driven PCI plan (ULP) This PCI plan aims to assign different DM RS base
sequences to neighbor cells, so that PUCCH interference due to DM RS collisions is
minimized. For this purpose, the network graph is divided into 30 subdomains so as to
10 20 km0
Fig. 4 Simulation scenario
R. Acedo-Hernandez et al.
123
avoid that PCIðcÞmod 30 is the same between neighbor cells. Thus, ULP achieves the
best PUCCH performance at the expense of deteriorating DL network perfor-
mance,since the avoidance of collision and confusions or DL CRS collisions is not
considered.
b. DL-driven PCI plan (DLP) In this plan, PCIs are assigned based only on DL
considerations, i.e., avoiding PCI collision and confusion and DL CRS collisions.
Since UL performance is not considered in this plan, DM RS collisions might occur in
PUCCH and PUCCH performance should be worse than with ULP.
c. Random PCI plan (RP) In this plan, used only for benchmarking purposes, PCI values
are randomly assigned to cells. In this work, RP is generated by building 100 random
PCI plans and selecting that providing the minimum number of collision and confusion
problems, as it is currently the criterion used by operators to build their PCI plans.
d. UL worst-case PCI plan (ULWP) In this plan, all neighbor cells have the same DM RS
base sequence, and thus collide in PUCCH. This case is used to derive the worst-case
PUCCH performance.
ULP and DLP are constructed by solving the graph coloring problem behind PCI planning
with the evolutionary multilevel graph partitioning algorithm described in [28]. In the
network graph, the vertices are the eNodeBs in the network, while the edges are the
adjacencies in the network. Edge weights reflect the penalty of having two neighbor cells
with the same PCI, DL RS or DM RS, as
cij ¼ xPCI � cPCIij þ xDLCRS � cDLRSij þ xDMRS � cDMRSij ; ð21Þ
where xPCI , xDLCRS and xDMRS are constant values depending on the selected PCI plan.
For the ULP plan, xPCI ¼ xDLCRS ¼ 0 and xPCI ¼ 1 (i.e., only DM RS collisions are
considered). For the DLP plan, xDMRS ¼ 0 and xPCI ¼ xDLCRS ¼ 1, i.e., only PCI
Table 2 Simulation parameters
Parameter Value
Simulator type System-level, static (grid-based)
PUCCH traffic model Poisson arrivals
Fixed service time 1 ms
Uniform /non-uniform spatial distribution
Propagation model Pathloss COST-231 Okumura Hata model
Shadowing lognormal fading (8 dB std., correlation distance 20 m)
Antenna model Antenna configuration SIMO (1x2)
PUCCH frame formats 1/1a/1b, 2/2a/2b
Open-loop/fast closed-loop power control, P0;PUCCH ¼ �120 dBm
UE model Transmit power 23 dBm
Antenna gain 0 dB
Antenna height 1.5 m
eNB model Antenna height 30 m
Max. antenna gain 17 dB
Noise floor �119.7 dBm
Interference model Orthogonality factors 2 ½0; 1�
Analysis of the Influence of PCI Planning on the…
123
confusion and CRS collision problems are considered. Once the network graph is defined,
the PCI planning problem aims to partition the graph into a fixed number of subdomains k,
V1;V2; . . .;Vk, so that the edge-cut is maximized. Specifically, k = 30 for ULP (DM RS
possible values) and for the DLP case, two partitions are done, first a partition of k = 3
groups to avoid equal PCIðcÞmod 3 (i.e DL CRS collisions) between adjacent cells and
then, in a graph with the same vertices but different edges, a partition of k = 168 groups.
The edges of this graph links adjacent neighbors and adjacent of adjacent neighbors trying
to avoid PCI collision and confusion problem. Then, the sum of both partition is the PCI
plan of 504 values, whose PCIðcÞmod 30 offers the DLP plan for DM RS signals.
For each of the 4 above-described PCI plans, 2 variants are simulated, considering that
sequence-group hopping is enabled or disabled. Thus, a total of 4 � 2 ¼ 8 different PCI
planning schemes are simulated per use case. When sequence-group hopping is enabled,
the PCI plan consists of groups of 30 cells sharing the same hopping-pattern, where a cell-
specific shift is added to avoid DM RS collisions between cells of the same cluster (i.e.,
Iinter(c, c0) = 0 8c, c0 in the same cluster) [18, 19].
Method assessment is based on the overall PUCCH outage probability, computed as the
average PUCCH outage probability across cells in the scenario. In this work, SINRth ¼�4:4 dB for type 1/1a/1b and �4:2 dB for type 2/2a/2b frame format [14].
The performance of each of the 8 PCI plans is evaluated in the 4 use cases for the 2
PUCCH frame formats (type 1/1a/1b vs 2/2a/2b) for different PUCCH traffic intensity
values with uniform and non-uniform spatial user distribution. In the uniform set-up, traffic
intensity is controlled by adjusting the PUCCH arrival rate through the parameter nr, i.e.,
the average number of PUCCH requests per PUCCH region and Transmission Time
Interval. For the uneven traffic distribution, the same parameter is computed on a cell basis
by multiplying the global PUCCH arrival rate in the network by the share of attempts per
cell, which is derived from PUCCH attempt statistics in the live network.
4.2 Analysis Results
For brevity, the analysis is focused on the results of format 1/1a/1b, and only a few cases
are shown for format 2/2a/2b.
4.2.1 Frame Format 1/1a/1b
The analysis is first focused on the results with ideal conditions and the impact of realistic
conditions is evaluated later.
a. Case 1 Ideal case (closed-loop PC, full orthogonality, uniform traffic).
Figure 5 shows the results for the 8 PCI planning schemes obtained from the 4 PCI
plans (ULP, DLP, RP and ULWP) with and without sequence-group hopping. Solid
lines represent methods with hopping while dotted lines correspond to methods
without hopping. The x-axis represent PUCCH traffic intensity in terms of the average
number of users per PUCCH region, Nus (recall that the maximum PUCCH capacity in
format 1/1a/1b is 18 users). The y-axis represents the average PUCCH outage in cells
of the scenario, PUCCHoutage.
As expected, it can be observed that the worst performance is achieved by ULWP
without hopping, where DM RS sequences are the same for all cells. This method is
the only one with non-negligible outage values, of up to 2.4%, in this scenario. Then, it
can be concluded that, in the ideal case, sequence-group and cyclic-shift hopping avoid
R. Acedo-Hernandez et al.
123
PUCCH failures regardless of the PCI plan. Likewise, without hopping, there is no
difference between planning PCI values randomly, based on UL criteria or based on
DL criteria.
b. Case 2 Non-ideal PC scheme.
Figure 6 shows the results of the different PCI planning schemes with open-loop
power control. In the figure, it is observed that all hopping schemes still have null
values of PUCCH outage probability with open-loop PC. However, the performance of
non-hopping schemes with high traffic is impaired by the use of open-loop PC. Thus,
outage probability with open-loop PC can be of up to 3.36% for ULWP, 0.09% for
DLP and 0.13% for RP, without hopping. ULP is the only method that achieves zero
outage probability with high traffic without hopping. Such a good performance comes
from the avoidance of UL DM RS collisions.
c. Case 3 Partial orthogonality between sequences.
Figure 7 presents the results of the PCI planning schemes with non-perfect
orthogonality between different cyclic shifts and base sequences. From the comparison
with previous figures, it can be deduced that the lack of orthogonality has a very
negative impact on PUCCH performance. Such an impairment limits the maximum
number of users transmitting in a PUCCH region at the same time. For instance, for a
maximum overall PUCCH outage probability of 0.1 , the maximum number of
simultaneous users per PUCCH region is 12 for all PCI planning schemes (ULP, DLP,
RP and ULWP) with hopping. Without hopping, this limit decreases to 5 users for
ULWP, 11 for RP and 12 for DLP and ULP case. Again, it is observed that DLP and
ULP have similar performance.
d. Case 4 Impact of uneven traffic.
Figure 8 shows the results of the PCI planning schemes with the uneven PUCCH
traffic distribution from the real network. When compared to the uniform case, it is
observed that uneven traffic distribution makes that the increase in PUCCH outage
probability with traffic intensity in most schemes is not so steep.
6 8 10 12 14 16 18 2010−3
10−2
10−1
100
101
nr
PUCCH
outage[%
]
ULP hoppingULP no hoppingDLP hoppingDLP no hoppingRP hoppingRP no hoppingULWP hoppingULWP no hopping
Fig. 5 PUCCH performance in the ideal case for frame format 1/1a/1b
Analysis of the Influence of PCI Planning on the…
123
4.2.2 Frame Format 2/2a/2b
The analysis for this format is restricted to case 4 (uneven traffic distribution), which is the
most realistic case. Recall that the maximum PUCCH capacity in format 2/2a/2b with the
considered settings is 6 users. Figure 9 shows that, in format 2/2a/2b, only ULW and RP
without hopping show unacceptable PUCCH performance with high traffic intensity.
6 8 10 12 14 16 18 2010−3
10−2
10−1
100
101
nr
PUCCH
outage[%
]
ULP hoppingULP no hoppingDLP hoppingDLP no hoppingRP hoppingRP no hoppingULWP hoppingULWP no hopping
Fig. 6 PUCCH performance with open-loop power control for frame format 1/1a/1b
2 4 6 8 10 12 14 16 18 2010−4
10−3
10−2
10−1
100
101
102
nr
PUCCH
outage[%
]
ULP hoppingULP no hoppingDLP hoppingDLP no hoppingRP hoppingRP no hoppingULWP hoppingULWP no hopping
Fig. 7 PUCCH performance with partial orthogonality for frame format 1/1a/1b
R. Acedo-Hernandez et al.
123
5 Conclusions
In this paper, a comprehensive performance analysis of the impact of PCI planning on the
performance of PUCCH in LTE has been carried out. For this purpose, a new analytical
model that reflects the influence of PCI planning on interference and outage probability due
2 4 6 8 10 12 14 16 18 2010−3
10−2
10−1
100
101
102
nr
PUCCH
outage[%
]
ULP hoppingULP no hoppingDLP hoppingDLP no hoppingRP hoppingRP no hoppingULWP hoppingULWP no hopping
Fig. 8 PUCCH performance with uneven traffic distribution for frame format 1/1a/1b
3 4 5 6 7 810−3
10−2
10−1
100
101
102
nr
PUCCH
outage[%
]
ULP hoppingULP no hoppingDLP hoppingDLP no hoppingRP hoppingRP no hoppingULWP hoppingULWP no hopping
Fig. 9 PUCCH performance with uneven traffic distribution for frame format 2/2a/2b
Analysis of the Influence of PCI Planning on the…
123
to DM RS collisions in PUCCH is presented. Such a model has been integrated into a static
system-level simulator to check the performance of several PCI planning schemes with
different PUCCH frame formats and network conditions.
Simulation results in a real scenario have shown that PCI planning has a significant
impact on PUCCH outage probability when base sequence hopping is disabled. The main
factor affecting PUCCH performance is non-perfect orthogonality between DM RS
sequences due to imperfect sequence design and multipath environment. Thus, improper
PCI planning can greatly reduce the maximum number of simultaneous PUCCH users,
especially when open-loop power control is used (e.g., in short data connections prevailing
in current LTE networks).
It has also been shown that a PCI plan only designed to avoid PCI collision/confusion
and reference signals collisions in the downlink achieves near-optimal performance in
terms of DM RS collisions in PUCCH. However, in extreme cases, DM RS collisions
caused by neighbor cells sharing the same sequence might degrade PUCCH performance
significantly. This is especially true in highly populated areas with small inter-site distance
if code protection is not effective.
Acknowledgements This work has been funded by the Spanish Ministry of Economy and Competitiveness(TEC2015-69982-R) and Ericsson, Agencia IDEA (Consejerıa de Ciencia, Innovacion y Empresa, Junta deAndalucıa, ref. 59288) and FEDER.
References
1. Acedo-Hernandez, R., Toril, M., Luna-Ramırez, S., de la Bandera, I., & Faour, N. (2015). Analysis ofthe impact of PCI planning on downlink throughput performance in LTE. Computer Networks, 76(1),42–54. doi:10.1016/j.comnet.2014.10.023.
2. Ahmed, F., Tirkkonen, O., Peltomki, M., Koljonen, J. M., Yu, C., & Alava, M. (2010). Distributedgraph coloring for self-organization in LTE networks. Journal of Electrical and Computer Engineering.doi:10.1155/2010/402831.
3. 3G Americas (2009). The benefits of son in lte: Self-optimizing and self-organizing networks. Tech. rep.4. Bandh, T., Carle, G., & Sanneck, H. (2009). Graph coloring based Physical-Cell-ID assignment for LTE
networks. In Wireless communications and mobile computing (pp. 116–120).5. Budisin, S. (2010). Decimation generator of Zadoff-Chu sequences. Sequences and Their Applications-
SETA, 6338, 30–40. doi:10.1007/978-3-642-15874-2.6. Burstrom, P., Falahati, S., & Simonsson, A. (2008). Uplink control channel in E-UTRA, radio link and
radio network evaluation. In Wireless communications and networking conference (pp. 835–839).doi:10.1109/WCNC.2008.153.
7. COST Action 231 (1999). Digital mobile radio. towards future generation system final report. Tech.Rep. EUR 18957, Ch. 4, European Communities
8. Dabak, A. G., Onggosanusi, E. N., & Papasakellariou, A. (2008). Computer generated sequences fordownlink and uplink signals in wireless communication systems.
9. Fernandez-Segovia, J. A., Luna-Ramırez, S., Toril, M., Vallejo-Mora, A., & Ubeda, C. (2015). Acomputationally efficient method for self-planning uplink power control parameters in LTE. EURASIPJournal on Wireless Communications and Networking. doi:10.1186/s13638-015-0320-7.
10. 3rd Generation Parthnership Project (2008). 3GPP TSG-RAN WG3 metting # 59, r3-080376, SON UseCase: Cell Phy_ID Automated Configuration. Tech. rep. www.3gpp.org (Accessed 3 June 2011).
11. 3rd Generation Parthnership Project (2008). 3GPP TSG-RAN WG3 metting # 59, R3-080812, solu-tion(s) to the 336.902 automated configuration of physical cell identity use case. Tech. rep., www.3gpp.org (Accessed 3 June).
12. 3rd Generation Parthnership Project (2009-10). Technical specification group radio access network;evolved universal terrestrial radio access network (e-utra); physical layer procedures (release 8), tr36.213 v8.8.0. Tech. rep.
R. Acedo-Hernandez et al.
123
13. 3rd Generation Parthnership Project (2011). Technical specification group radio access network;evolved universal terrestrial radio access network (e-utran); self-configuring and self-optimizing net-work (son) use cases and solutions (release 9), tr 36.902 v9.3.1. Tech. rep.
14. 3rd Generation Parthnership Project (2013). Lte; evolved universal terrestrial radio access (e-utra);relay radio transmission and reception, ts 36.116. Tech. Rep. Release 11.
15. 3rd Generation Parthnership Project (2013). Technical specification group radio access network;evolved universal terrestrial radio access network (e-utra); physical channel and modulation (release11), tr 36.211 v11.2.0. Tech. rep.
16. Gupta, A., & Jha, R. K. (2015). A survey of 5g network: Architecture and emerging technologies. IEEEAccess, 3, 1206–1232. doi:10.1109/ACCESS.2015.2461602.
17. Hamalainen, S., Sanneck, H., & Sartori, C. (2012). LTE self-organising networks (SON): Networkmanagement automation for operational efficiency. Hoboken: Wiley.
18. Holma, H., & Toskala, A. (2011). LTE for UMTS: Evolution to LTE-advanced. Hoboken: Wiley.19. Kavlak, H., & Ilk, H. (2012). PCI planning strategies for long term evolution networks. In International
Conference on Networking, IFIP (pp. 151–156).20. Krichen, M., Barth, D., & OMarce (2012). Performances evaluation of different algorithms for PCIs self
configuration in LTE. In 18th IEEE international conference on networks (ICON). doi:10.1109/ICON.2012.6506558.
21. Lim, J., & Hong, D. (2011). Management of neighbor cell lists and physical cell identifiers in self-organizing heterogeneous networks. Communications and Networks, 13, 367–376.
22. Liu, Y., Li, W., Zhang, H., & Lu, W. (2010). Graph based automatic centralized PCI assignment inLTE. In IEEE symposium on computers and communications (ISCC) (pp. 919–921). doi:10.1109/ISCC.2010.5546812.
23. Liu, Y., Li, W., Zhang, H., & LYu (2010). Distributed PCI assignment in LTE based on consultationmechanism. In 6th international conference on wireless communications networking and mobilecomputing (WiCOM) (pp. 1–4). doi:10.1109/WICOM.2010.5601210.
24. NGMN (2006). Next generation mobile networks beyond hspa & evdo v3.0. Tech. rep.25. Oppolzer, J., & Bestak, R. (2012). Physical cell identifier assignment in LTE-advanced networks. In 5th
joint IFIP wireless and mobile networking conference (WMNC). doi:10.1109/WMNC.2012.6416150.26. Ramiro, J. (2011). Self-organizing networks: Self-planning, self-optimization and self-healing for GSM,
UMTS and LTE. Hoboken: Wiley.27. Sesia, S., Toufik, I., & Baker, M. (2009). LTE: The UMTS long term evolution, from theory to practice.
Hoboken: Wiley.28. Soper, A., Walshaw, C., & Cross, M. (2004). A combined evolutionary search and multilevel opti-
misation approach to graph-partitioning. Journal of Global Optimization, 29, 225–241.29. Vadada, H. (2011). LTE PCI planning. Retrieved September 17, 2015, from: http://www.telecom-cloud.
net/wp-content/uploads/2010/09/PCI-Planning-for-LTE.pdf.30. Wei, Y., Peng, M., Wang, W., Min, S., Jiang, J. M., & Huang, Y. (2012). Automatic distributing
schemes of physical cell identity for self-organizing networks. International Journal of DistributedSensor Networks. doi:10.1155/2012/973713.
31. Wu, T., Rui, L., Xiong, A., & Guo, S. (2010). An automation PCI allocation method for enodeB andhome enodeB cell. In 6th international conference on wireless communications networking and mobilecomputing (WiCOM). doi:10.1109/WICOM.2010.5600764.
32. Xu, H., Zhou, X.W., & Li, Y. (2011). Model of hypergraph colouring for Self-configuration in LTEnetworks. In International conference on information management, innovation management andindustrial engineering (ICIII) (Vol. 1). doi:10.1109/ICIII.2011.100.
33. Zahran, A. (2012). Extended synchronization signals for eliminating PCI confusion in heterogeneousLTE. In IEEE Wireless Communications and Networking Conference (WCNC) (pp. 2588–2592). doi:10.1109/WCNC.2012.6214236.
34. Zhang, X., Zhou, D., Xiao, Z., Liu, E., Zhang, J., & Glasunov, A. (2011). Dynamic group PCIassignment scheme. In The 7th int. conference on wireless and mobile communications, ICWMC (pp.101–106).
35. Zyren, J., & McCoy, D. (2007). Overview of the 3g pp long term evolution physical layer. Tech. rep.,Freescale Semiconductor
Analysis of the Influence of PCI Planning on the…
123
R. Acedo-Hernandez received the MSc degree in TelecommunicationEngineering in 2014 from the University of Malaga, Spain. In 2012,she joined the Communications Engineering Department, University ofMalaga, where she is currently working toward the Ph.D. degree intelecommunications engineering in a collaborative project with Eric-sson. Her research interests are focused on self-planning and self-optimization of radio access networks and graph partitioning.
M. Toril gained his M.S. in Telecommunication Engineering andPh.D. degrees from the University of Malaga, Spain, in 1995 and 2007,respectively. Since 1997, he is Lecturer in the CommunicationsEngineering Department, University of Malaga, where he is currentlyAssociate Professor. He has authored more than 90 publications inleading conferences and journals and 3 patents owned by Nokia Cor-poration. His current research interests include self-organizing net-works, radio resource management and graph partitioning.
S. Luna-Ramırez received the M.S. degree in telecommunicationengineering and the Ph.D. degree, both from the University of Malaga,Malaga, Spain, in 2000 and 2010, respectively. Since 2000, he hasbeen with the Department of Communications Engineering, Universityof Malaga, where he is currently Associate Professor. His researchinterests include self-optimisation of mobile radio access networks andradio resource management.
R. Acedo-Hernandez et al.
123
J. A. Fernandez-Segovia gained his M.S. in TelecommunicationEngineering and Ph.D. degrees, both from the University of Malaga(Spain), in 2010 and 2015, respectively. From 2010 to 2012 he wasworking in GSM/UMTS networks O&M projects as external con-tractor in Vodafone Spain. Since 2012, he has been with the Depart-ment of Communications Engineering, University of Malaga, andOptimi-Ericsson, where he is currently research assistant. His researchinterests include self-planning and optimization of mobile networks.
C. Ubeda received his MSc degree in Telecommunication Engineeringfrom Miguel Hernandez University of Elche in 2006. After finishinghis Master’s Thesis at Aalborg University, he worked as externalresearcher for Nokia Siemens Networks in Aalborg. In 2008 he joinedOptimi as R&D engineer in Malaga. Since 2011 he is a senior researchengineer for Ericsson in Madrid. His main research interests includeradio resource management, RF planning and mobile networkoptimization.
Analysis of the Influence of PCI Planning on the…
123